Abstract
In recent years, reliance on data-driven adaptive learning technologies has increased substantially within virtual learning environments. However, their application within gamified learning environments may introduce unintended bias in the interpretation of learner engagement. This study investigates how fairness-aware learning analytics can inform adaptive gamification to support inclusive and sustainable virtual learning. The Open University Learning Analytic Dataset (OULAD) was adopted and used in analysing behavioural, demographic, and assessment data for students with or without disabilities. This was with the purpose of examining patterns of engagement and predicting academic achievement among these groups of students. Both logistic regression and gradient boosting predictive models were employed and evaluated using performance metrics and fairness measures. The results show that students with disabilities recorded lower levels of early engagement, while the logistic regression model achieved high overall accuracy but produced inconsistent outcomes across learner groups. Findings also reveal that the application of fairness-aware threshold calibration brought about reduction in group-level differences with sustained predictive performance. These findings indicate the potential for integrating fairness-aware analytics into adaptive learning systems for the purpose of supporting balanced engagement and promoting inclusive gamification. This study therefore provides actionable guidance for the development of ethical and sustainable data-driven learning technologies.